After the introduction of the ultrasound into the clinical evaluation of thyroid diseases in the late 1960s, the frequency of detected thyroid nodules has been increased due to the widespread use of ultrasonography [1, 2]. Ultrasound-guided FNA cytology serves to provide the most appropriate clinical management of thyroid nodules. TBSRTC including six categories with suggestions of appropriate clinical management for each category provides a standardized reporting format for thyroid FNAs which has been updated with significant developments during the last two decades following its first edition [6]. However, indeterminate categories continue to exist including probably the most problematic category “Atypia of Undetermined Significance (Bethesda III)” in the 3rd edition of the reporting system [6]. In this context, the present study aimed to investigate the impact of AUS subcategories and subgroups designed according to the criteria reported in the 2nd [9] and in the 3rd edition [6] of TBSRTC on ROM via a three-staged study plan. In the first stage, nuclear and architectural features were evaluated, and these features were compared with the histopathological diagnostic groups determined according to the diagnostic categories and tumor types reported in the 5th edition (2022 Beta version) of the WHO Classification of Endocrine and Neuroendocrine Tumours [28], and comparisons of ROM were performed between all of the features. In the second stage, the current study tried to develop a nuclear scoring schema inspired by the study of Nikiforov et al. [30]. Finally, in the third stage, subcategories and subgroups were created based on NS, other cellular features, and architectural features, and these subcategories and subgroups were compared in terms of histopathological diagnosis and ROM. So, the most remarkable results of the present study can be summarized as in the following: (i) The umbrella term as AUS-nuclear subcategory, particularly subgroups including focal marked nuclear atypia (focal NS 3) and nuclear atypia with architectural atypia, constitutes a high-risk group in terms of malignancy; (ii) nuclear features including nuclear grooves, nuclear overlapping, molding, nuclear contour irregularity, chromatin clearing, and margination are independent significant predictors of malignancy; (iii) three-tiered nuclear scoring scheme may provide a more objective and reproducible assessment method in the evaluation of thyroid FNAs as in the assessment of histological nuclear features.

Although 3rd edition of TBSRTC [6] continues to suggest an upper limit as 10% for AUS in all thyroid FNAs, the reported rates range between 1 [10] and 20% [23] (Table 13). The rate of AUS was 17.6% in the current study (This value reflects the general data in our department. It does not include the evaluations made in the present study.), and this value was higher than the proposed rate of AUS in TBSRTC 3rd edition [6]. Probable cause of the higher rate of AUS may be the subjective application of the defined objective criteria in the evaluation of thyroid FNAs. Also, in each edition, the rate of ROM was updated according to the accumulated data and the effects of changing terminologies [6]. However, the reported rates of ROM in the literature have been often higher than the recommended values during the age of TBSRTC (Table 13). The ROM for categories was reported as mean and expected ranges in the recent edition of TBSRTC and ROM for AUS was reported as 22% (13–30) [6]. The current gold standard calculation of ROM is obtained by dividing the total number of histopathologically confirmed malignant cases by the AUS cases with surgical follow-ups; however, the calculation method is controversial since approximately half of the thyroid nodules in the AUS category do not have surgical follow-ups. Most of the data reported about the rates of ROM in the literature are the results of the studies performed before the era of NIFTP and ranges between 17.0 [15] and 83.1% [16]. Liu et al. [8] excluded NIFTP from the malignant diagnostic category and informed ROM as 54.3% in their study. In the present study, histological diagnostic categories were created according to the 5th edition (Beta version) of the WHO Classification of Endocrine and Neuroendocrine Tumours [28], and NIFTP was categorized as a LRN and was not considered a malignant tumor in the calculation of ROM. The risk of malignancy was 30.7% in the present study and was slightly higher than the upper value of the expected range in the 3rd edition of TBSRTC [6] but lower than the rate reported by Liu et al. [8]. The higher rate of ROM in the present study may be due to the selection of high-risk nodules for surgery by evaluating patients diagnosed as AUS with clinical and radiological findings in a multidisciplinary manner. So, wide range between the reported results for the rates of AUS and ROM, which mostly represent real-life data, may be related to various factors, including inadequate clinicopathological communication, insufficient practice of objective criteria for evaluation of FNAs, the tendency of observers to stay in the safe zone, and the lack of experience of researchers.

The present study investigated the impact of nuclear and architectural features on risk of malignancy in FNAs with AUS. All of patients in the study group had surgical follow-ups, and the FNA localization and the target lesion in the surgical specimen overlapped. The first FNAs of each nodule were included in the study in order to prevent previous FNA-related regenerative changes from affecting the evaluation. Microfollicular pattern and trabecular pattern were seen significantly more frequently in FNAs of BNs and LRNs, while three-dimensional groups were significantly more common in FNAs of MNs in surgical specimens. And also, three-dimensional groups show the highest ROM value with close to significance. Gularia et al. [31] and Kaymaz et al. [32] showed that cases containing a three-dimensional group had higher ROM values. So, a more skeptical investigation of FNAs with three-dimensional groups for an additional evidence of malignancy may be argued. Additionally, definition of streaming pattern which was detected in 44 cases in our study is reported for the first time in the literature. Although, a partially similar pattern is mentioned in the TBSRTC 3rd edition (architectural atypia, criteria 3) [6], it is not thought to fully correspond to the pattern described here. When we examined the cases with streaming patterns, it was determined that initial FNAs of BNs more frequently exhibited this pattern without statistically significance. However, considering that all of the cases in our study were in the AUS category, our findings may not clearly reflect the nature of the streaming pattern, and investigation of this pattern in a study group with a larger number of cases including all of the TBSRTC categories may reveal the impact of this pattern on thyroid cytology.

Comparisons of nuclear features with ROM values revealed that the risk of malignancy in FNAs with nuclear grooves, chromatin margination, nuclear molding, nuclear contour irregularity, and nuclear overlapping (in an order with descending ROM values) was significantly higher than in FNAs without these features. Also, nuclear grooves, nuclear overlapping, molding, nuclear contour irregularity, chromatin clearing, and chromatin margination (in an order with descending Odds ratios) were significant independent predictors of histological malignancy in multivariate analysis. However, nuclear overlapping, chromatin clearing, and chromatin margination were also more common in the group consisting of histological LRNs. There was no relationship between nuclear elongation and nuclear enlargement with histological malignancy. Kato et al. [33] reported that the presence of nuclear grooves may indicate histological malignancy in indeterminate thyroid cytology. Additionally, the authors stated that the presence of four or more atypical nuclear features or coexistence of nuclear grooves and inclusions may be associated with malignancy. Kaymaz et al. stated that membrane irregularities such as pseudoinclusion, nuclear contour irregularity, and nuclear grooves were associated with malignancy, and also, nuclear elongation and overlapping were predictive for malignancy [32]. FNAs with nuclear grooves revealed the highest risk of malignancy, and in multivariate analysis, the presence of nuclear grooves resulted as an independent significant predictor for malignancy with the highest value of odds ratio as the authors have reported. On the other hand, the presence of more than two of six atypical nuclear features was found to be a significant likelihood of malignancy in the ROC analysis performed in this study. The previous results reported in terms of the relationship between malignancy with nuclear contour irregularity, nuclear overlapping, and nuclear grooves were similar with the results of the present study regarding these parameters. However, according to the definition in TBSRTC [6, 9], pseudoinclusion should by definition not be present in the evaluation of nuclear atypia for AUS except atypical cyst-lining cells. In the current study, there was no FNA revealing intranuclear pseudoinclusion in the study group. As a result, the presence of nuclear features investigated in the current study may be valuable criteria for marked nuclear atypia. Also, the presence of nuclear grooves or the presence of more than two of the mentioned atypical nuclear features may represent likelihood of malignancy.

The difficulty in using subcategories, as well as the difficulty in using the diagnostic category of AUS, is evidenced by the wide range of ROM reported in the previous studies (Table 13). These differences may probably result from changes in the identification of variables and the entities over time. However, the striking point is the presence of interobserver variability in the evaluation of cytological atypia despite the reported widespread definitions and the criteria. Although the features called cytological or nuclear atypia (such as nuclear enlargement, chromatin clearing, and nuclear membrane irregularity) are the same among cytopathologists, there is significant subjectivity in the evaluation and standardization is quite weak. In the recent history of endocrine pathology, definition of nuclear scoring for nuclear features has made a significant contribution to the evaluation of encapsulated follicular patterned thyroid tumors with papillary-like nuclear features [30]. So, the current study tried to develop a nuclear scoring schema inspired by the study of Nikiforov et al. [30] in order to reduce this subjectivity in the evaluation of nuclear features. Thus, nuclear features were evaluated under the headings of (1) size and shape, (2) nuclear membrane irregularities, and (3) chromatin features resulting in a NS ranging from 0 to 3. Among the NS groups, the group with the highest ROM with a rate of 64.2% (n = 34) is the NS3 group. Compared to the overall ROM in the present study (30.7%), the malignancy risk of focal/diffuse NS3 was above this rate and was significantly higher than the other NSs. In this context, focal or diffuse NS3 was found to be more noteworthy in cytological evaluation. The NS2 was significantly more common in FNAs of BNs. Guleria et al. [31] and Kaymaz et al. [32] used similar nuclear scoring models as in the current study and showed that the NS2-3 group was more frequently associated with malignancy. However, in these studies, ROM of NSs were not presented separately, and their focal/diffuse status was not evaluated. Altınboğa et al. [34] examined aspirates with AUS according to the NS, but the scoring system they used was different from the scoring system used in the present study and was designed according to the percentages of nuclear features, and scores were given between 0 and 10 points for nuclear features. For this reason, the relationship between NS and ROM could not be compared with the results reported by the authors. However, there are also studies examining the cytological features of NIFTP using similar scoring system [29, 35, 36]. According to the data of these studies, which evaluated cytological features in TBSRTC indeterminate categories, it can be inferred that cases with NS2-3 have a higher ROM compared to cases with NS0-1. Evaluation of nuclear features by a scoring schema may provide a more objective and safer evaluation method in the categorization and subcategorization of thyroid FNAs, as it can transform the detected findings into numerical data.

The TBSRTC categorization and reported recommendations for each category standardizes the reporting of thyroid cytology and the clinical management of thyroid nodules. However, the criteria that will be reflected in the microscope objective for each category are defined in detail; the interpretation of what is reflected from the microscope eyepieces varies among pathologists. The fact that most of the studies published since the first edition reported values above the recommended rates of AUS and ROM by TBSRTC is a reflection of these variabilities. In the current study, which was carried out during the transition zone between the TBSRTC 2nd edition [9] and TBSRTC 3rd edition [6], 6 subcategories and 11 subgroups [AUS-N (AUS-N1-4), AUS-A (A1-2), AUS-N&A, AUS-O (O1-2), AUS-NOS (NOS1-3), and AUS-L] were created based on the definitions and criteria reported for each category in the TBSRTC 2nd edition [9], updated according to according to the reported terms in the 3rd edition of TBSRTC [6] and powered by the nuclear scoring schema, in order to contribute to a more objective evaluation. So, the definitions of subcategories and subgroups in the present study enclosed the successor and predecessor criteria of TBSRTC (Table 2) [6, 9].

Accumulated data from published studies reporting ROM for AUS following the 1st edition [7] resulted in subcategories as AUS-nuclear and AUS-other in the 3rd edition [6]. The subcategories reported in some of the studies on this subject in the literature and the malignancy rates evaluated in surgical resection materials for each category are summarized in Table 13. In these studies, there are different terminologies that may cover the AUS-nuclear subcategory and subcategories with different numbers and definitions that may correspond to the AUS-other subcategory. In subcategories that may include the AUS-nuclear subcategory, ROM varies between 28 [21] and 100% [26]. While some of these mentioned studies evaluated architectural atypia separately, some examined it within other subcategories. Additionally, reported ROM values for architectural atypia vary between 6.9 [21] and 73.3% [37].

In this study, the AUS-nuclear subcategory covers the cytological atypia criteria in the TBSRTC 2nd edition [9]; it includes definitions other than the nuclear and architectural atypia criteria defined under the AUS-nuclear atypia subcategory in the TBSRTC 3rd edition [6]. ROM of the AUS-N subcategory was significantly higher than the other subcategories (AUS-L subcategory was ignored due to the low number of cases). Therefore, if the results of the current study and the data accumulated from previous studies are combined, the superiority of the AUS-N subcategory over other subcategories in predicting histological malignancy is clear, and the subcategorization reported in the TBSRTC 3rd edition [6] may be foreseeable to benefit in the clinical management of thyroid nodules diagnosed with AUS. In this case, the question may emerge: Does every nuclear atypia express this prediction correctly? In order to clarify this question, the AUS-N subcategory was examined separately in four subgroups in this study. AUS-N1 subgroup had higher ROM than other AUS-N subgroups and all other subcategories and subgroups in the study. In other words, in FNAs with focal NS3 features, ROM is high, and this value was found to be 65.2% in this study. This value is close to the ranges [74 (67–83%)] reported in the TBSRTC 3rd edition [6] for “Suspicious for malignancy, Bethesda Category V.” Perhaps, it may be questioned that upgrading of FNAs with NS 3/AUS-N1 subgroup as SFM rather than the AUS category and the use of the AUS diagnostic category may be reduced. Therefore, ROM values were compared between AUS and SFM categories in the current study (ROM value for SFM category was obtained from the archive records without reevaluation of aspirates). ROM of SFM (95.1%) was significantly higher than ROM values for AUS (overall), AUS-nuclear subcategory, AUS-other subcategory, and AUS-NS3 subgroup. The higher ROM of SFM in our department was approximately within the recommended rates for Bethesda category VI, malignant [97% (97–100)] [6]. Lack of the minimum quantitative threshold for a diagnosis of malignancy in FNAs and also protective effect of the term as “suspicious” may be the causes of the higher rate of ROM in SFM. The quality (presence/absence of intranuclear pseudoinclusions) and quantity (extent of intranuclear pseudoinclusions for malignancy) of the nuclear features and the perception of these features on the pathologists may cause the same features to be evaluated in different TBSRTC categories by different pathologists. Although presence of a few intranuclear pseudoinclusions and/or extensive nuclear grooves in follicular cells and widespread of other atypical nuclear features (approximately > 70% of aspirates) was regarded as SFM rather than AUS in the present study, presented ROM for SFM obtained from the archive records may reflect the interobserver variability in indeterminate TBSRTC categories. Thus, studies including comparisons between the subgroups of the AUS-N subcategory with SFM category via a nuclear scoring schema in larger study groups may be beneficial in this regard.

Since, ROM is similar in aspirates with both mild nuclear and architectural atypia regardless of the presence or absence of concomitant architectural atypia, this subgroup is considered AUS-nuclear subcategory in the TBSRTC 3rd edition [6]. AUS-N&A subcategory had the third highest ROM rate following the AUS-N1 subgroup and AUS-N subcategory in the current study. The rate of LRN was also high in the surgical materials of the AUS-N&A subcategory. Therefore, we examined the NS distribution of AUS-N&A subcategory. These evaluations showed that the AUS-N&A subcategory with NS3 features had significantly higher ROM (60.0%) than AUS-N&A aspirates evaluated as NS2 (28.1%) (p = 0.006). Also, ROM for extensive NS2 features (40.6%) was higher than the focal NS2 features (15.6%) with close to the significance (p = 0.052). These findings emphasize the importance of nuclear scoring in thyroid FNA samples. AUS-A subcategory and AUS-A2 subgroup significantly indicated BNs on histological examination. Considering the relationship of three-dimensional groups with high ROM, which is emphasized in some previous studies [31, 32] and in the current study, perhaps architectural atypia, as well as nuclear atypia, may be nature-defining. So, detailed definition of architectural features as well as nuclear scoring and raising awareness in terms of three-dimensional groups may be useful in the correct evaluation of thyroid cytology.

On the other hand, a significantly higher rate of NND was detected in the surgical specimens of nodules considered AUS-O and AUS-NOS subcategories with the lowest ROM value in the AUS-O category (5.6%) which was followed by AUS-NOS (11.5%). The ROM value determined for the AUS-O subcategory in this study is lower than the values reported in previous studies [16, 37,38,39] and the mean value reported in the TBSRTC 3rd edition [6]. A similar situation existed for the AUS-NOS subcategory, undoubtedly the category with the most different definitions [6, 16,17,18,19, 21, 31, 32, 39,40,41,42]. Park et al. [10, 31] used similar criteria for AUS-NOS as our study and detected a ROM of 14.5%, and this rate supports our study. Since clinical, radiological, and laboratory findings are important in cytological evaluations for the AUS-O and AUS-NOS subcategories, it may be due to the fact that evaluation of the study group in the multidisciplinary endocrine diseases council contributed to the differential diagnosis of these subcategories and to the selection of appropriate patients for surgery in the present study. Therefore, the integration of clinical and laboratory findings with cytological findings may save the AUS-NOS subcategory from being a wastebasket category.

In this study, all categories except the AUS-L subcategory were compared pairwise in terms of ROM. No statistically significant difference was observed between AUS-N and AUS-N&A subcategories in terms of ROM. In other words, these two subcategories exhibited similar patterns in terms of ROM. However, the ROM values of these two categories were found to be significantly higher than the ROM values of other subcategories. Therefore, these two subcategories were combined to form a group, and all other subcategories were combined to form a separate group. And finally, the ROM value of the group consisting of AUS-N and AUS-N&A subcategories was found to be significantly higher than the other group. Based on these results, the group consisting of AUS-N and AUS-N&A subcategories named as “high-risk group,” and the group consistent of other subcategories named as “low-risk group.” These last groups created actually correspond to the AUS-nuclear and AUS-other subcategories expressed in the TBSRTC 3rd edition [6]. Studies reported following the first introduction of TBSRTC [7] and a recently reported study based on TBSRTC 3rd edition [6], comparing the changes reported as AUS-nuclear or AUS-cytological atypia with other AUS defining changes, individually or in different combinations are summarized in Table 13 [8, 15,16,17,18, 20,21,22,23, 25, 26, 37,38,39, 43,